Cancer surgery is associated with high operative mortality rates, particularly at hospitals with low procedure volumes. Efforts to improve quality at these centers are hindered by lack of understanding about mechanisms responsible for observed volume-outcome relationships. This study will approach this question in 3 steps: 1. Determine what's different between high volume and low volume hospitals. Using 100% national samples from the 1994-2001 Medicare database (n approximately 800,000), we will compare structure and process variables at high and low volume hospitals with each of 8 procedures. Structural variables specific to cancer care will include on-site availability of radiation and chemotherapy services, hospital participation in cancer trials, and surgeon "specialization" in cancer surgery. We will then assess variables related to patient selection and other processes of perioperative care. 2. Identify, which factors help explain observed volume-outcome relationships. We will then examine relationships between structure, process, and outcomes (operative mortality) with the 8 procedures. Using nested models, we will assess how observed volume-outcome relationships change as various structural and process of care variables are added to the model. We will repeat these analyses using the SEER-Medicare linked database to better control for case-mix (tumor stage). 3. Prepare the groundwork for studying the fine details of care. To best plan a study of structure, process, and outcomes based on medical records or site visits, we must first understand whether variation in operative mortality rates across hospitals is attributable to specific causes of death. We will develop and test a chart-based instrument for assessing cause of death in a pilot study at two hospitals in northern New England.